mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-05 06:01:23 +00:00
[CK-Tile] Add the API to load SGPR (#2878)
* Have a workable version for SGPR * have a workable version for atomic add * Revert "have a workable version for atomic add" This reverts commit 792377a590c26cfff9c8f545d9a9e8484a7422eb. * substitute with the new sgpr read api * update the CHANGELOG * have a workable version for atomic add * Revert "have a workable version for atomic add" This reverts commit 792377a590c26cfff9c8f545d9a9e8484a7422eb. * change to static for logic * have a workable version for atomic add * Revert "have a workable version for atomic add" This reverts commit 792377a590c26cfff9c8f545d9a9e8484a7422eb.
This commit is contained in:
@@ -320,9 +320,9 @@ struct BlockFmhaFwdPagedKVPipelineQRKSVS
|
||||
k_block_tile = load_tile(k_dram_window);
|
||||
}
|
||||
auto physical_next_block_id_k =
|
||||
__builtin_amdgcn_readfirstlane(k_page_block_navigator.prefetch_table_id(
|
||||
amd_wave_read_first_lane(k_page_block_navigator.prefetch_table_id(
|
||||
i_page_block_k, k_dram_block_window, {kN0, 0}));
|
||||
auto physical_next_block_id_v = __builtin_amdgcn_readfirstlane(
|
||||
auto physical_next_block_id_v = amd_wave_read_first_lane(
|
||||
v_page_block_navigator.prefetch_table_id(i_page_block_v, v_dram_window, {0, kK1}));
|
||||
|
||||
if constexpr(BiasEnum == BlockAttentionBiasEnum::ELEMENTWISE_BIAS)
|
||||
|
||||
@@ -321,9 +321,9 @@ struct BlockFmhaFwdSplitKVPipelineQRKSVS
|
||||
k_block_tile = load_tile(k_dram_window);
|
||||
}
|
||||
auto physical_next_block_id_k =
|
||||
__builtin_amdgcn_readfirstlane(k_page_block_navigator.prefetch_table_id(
|
||||
amd_wave_read_first_lane(k_page_block_navigator.prefetch_table_id(
|
||||
i_page_block_k, k_dram_block_window, {kN0, 0}));
|
||||
auto physical_next_block_id_v = __builtin_amdgcn_readfirstlane(
|
||||
auto physical_next_block_id_v = amd_wave_read_first_lane(
|
||||
v_page_block_navigator.prefetch_table_id(i_page_block_v, v_dram_window, {0, kK1}));
|
||||
|
||||
if constexpr(BiasEnum == BlockAttentionBiasEnum::ELEMENTWISE_BIAS)
|
||||
@@ -618,7 +618,7 @@ struct BlockFmhaFwdSplitKVPipelineQRKSVS
|
||||
&i_page_block_v_ = i_page_block_v,
|
||||
&v_dram_window_ = v_dram_window](auto i_k1) {
|
||||
auto physical_next_block_id_v_ =
|
||||
__builtin_amdgcn_readfirstlane(v_page_block_navigator.prefetch_table_id(
|
||||
amd_wave_read_first_lane(v_page_block_navigator.prefetch_table_id(
|
||||
i_page_block_v_, v_dram_window_, {0, kK1}));
|
||||
const auto v = load_tile(v_dram_window_); // load next v
|
||||
block_sync_lds();
|
||||
|
||||
Reference in New Issue
Block a user